CN113011669B - Method for predicting monthly stock quantity of live pigs - Google Patents
Method for predicting monthly stock quantity of live pigs Download PDFInfo
- Publication number
- CN113011669B CN113011669B CN202110339114.9A CN202110339114A CN113011669B CN 113011669 B CN113011669 B CN 113011669B CN 202110339114 A CN202110339114 A CN 202110339114A CN 113011669 B CN113011669 B CN 113011669B
- Authority
- CN
- China
- Prior art keywords
- mode
- algorithm
- decomposition
- predicted value
- modes
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 241000282887 Suidae Species 0.000 title claims abstract description 26
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 19
- 238000012545 processing Methods 0.000 claims abstract description 14
- 238000010606 normalization Methods 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims abstract description 10
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000000694 effects Effects 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 7
- 238000009499 grossing Methods 0.000 claims description 6
- 238000001228 spectrum Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000009792 diffusion process Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 abstract 1
- 239000010410 layer Substances 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 7
- 230000003044 adaptive effect Effects 0.000 description 3
- 230000035945 sensitivity Effects 0.000 description 3
- 230000002411 adverse Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000013179 statistical model Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 241001465754 Metazoa Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/02—Agriculture; Fishing; Forestry; Mining
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Marine Sciences & Fisheries (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Animal Husbandry (AREA)
- Educational Administration (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Mining & Mineral Resources (AREA)
- Primary Health Care (AREA)
- Artificial Intelligence (AREA)
- Agronomy & Crop Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for predicting monthly stock quantity of live pigs, which comprises the following steps: decomposing original time sequence data x (t) of the live pig stock volume into K modes by a variational modal decomposition algorithm; for each mode obtained by the variational mode decomposition, normalization processing is carried out, and then the modes are divided into a training set and a test set according to a preset proportion; training an extreme learning machine algorithm by using training set data, and determining optimal parameters of the algorithm; selecting input set data of an extreme learning machine algorithm by using a sliding window with the step length of V; for each mode, inputting the input set data of the test set into the trained extreme learning machine algorithm, outputting the predicted value of the next time point, and performing reverse normalization processing on the predicted value to obtain a predicted value sequence u k (t); and adding and reconstructing the predicted values of all the modes to obtain a final predicted value result. The method can improve the prediction efficiency and the prediction precision, and verifies the prediction effectiveness in experiments.
Description
Technical Field
The invention relates to the technical field of prediction of monthly stock keeping quantity of live pigs, in particular to a method for predicting monthly stock keeping quantity of live pigs based on self-adaptive variational modal decomposition and an extreme learning machine.
Background
The prediction of data at a future time point by mining the intrinsic statistical characteristics of time series data currently exists in many scientific fields. Time series data consists of a series of observations taken at different time points. The purpose of mining the internal statistical characteristics of the time sequence data is to find out the change rule among historical sample point data and construct a time sequence model to carry out the out-of-sample prediction on the data at the future time point. Generally, data at different time points in a time series have the characteristic of consistent calibers. The time sequence data can be time point data or time period data according to different interception modes of the data.
The monthly stock quantity of domestic pigs is accurately predicted, and important reference is provided for formulating stage animal husbandry policies. However, under the combined action of various factors such as economy, environment and the like, the monthly stock-keeping quantity sequence of the live pigs is generally non-stable and non-linear, and the two characteristics provide higher requirements for accurate prediction of the live pigs. The methods widely used for predicting time series data at present can be mainly classified into three types: a metering statistical model, a prediction model based on an artificial intelligence method and a hybrid prediction model. The metric statistical models mainly include self-differential integrated moving average Autoregressive (ARIMA) models, Generalized Autoregressive conditional heterology (GARCH) models, and the like, and the models have good effects when processing linear characteristics of time series data, but often have poor performance when processing non-stationary and non-linear complex series data. In recent years, a prediction model based on Artificial intelligence is gradually applied to processing of complex time series data, and typically, an Artificial Neural Network (ANN), a Support Vector Machine (SVM), an Extreme Learning Machine (ELM), and the like exist.
However, the traditional ANN adjusts the weight parameters by using an iterative algorithm with gradient descent, which is usually accompanied by slow training speed, and is prone to fall into local optimization and overfitting, resulting in poor robustness of the prediction result. Compared with the ANN, the training speed and the robustness of the prediction result of the SVR are effectively improved, but the generalization capability of the SVR has higher sensitivity to the selection of the kernel function and the parameters thereof, and the prediction result with higher precision can be realized only by continuously adjusting the parameters. Compared with ANN and SVR, ELM is a single-layer forward feedback neural network learning algorithm, has the advantages of high operation speed and strong generalization capability, greatly improves the operation efficiency and convergence speed, and enables the prediction result to be more robust. At present, ELM shows good performance in the prediction of complex time series data.
The hybrid prediction model has multiple implementation forms, and at least two algorithms are combined together by the hybrid prediction model to achieve a better prediction effect. On one hand, the hybrid prediction model can combine a plurality of prediction models together, extract the characteristics of the complex time sequence data from different angles and respectively make predictions, and combine prediction results of various places to obtain a final prediction result; on the other hand, the hybrid prediction model can combine a data preprocessing algorithm with a prediction model, and the model firstly carries out preprocessing such as dimensionality reduction and decomposition on complex original data based on the concept of 'divide and conquer', carries out modeling prediction on the processed data, and then reconstructs a prediction result to obtain a final prediction result. Currently, common decomposition algorithms include wavelet transformation, Empirical Mode Decomposition (EMD), integrated Empirical mode decomposition (EEMD), adaptive noise complete integration Empirical mode decomposition (CEEMDAN), modified CEEMDAN (Improved complex Empirical mode decomposition with adaptive noise, icemdan), Variable Mode Decomposition (VMD), and the like.
The EMD is a signal time-frequency analysis method, can realize the stabilization of non-stationary nonlinear data, and plays a great role in the analysis of financial time-series data. However, this method also has some drawbacks, mainly mode boundary effects, mode overlap, sensitivity to noise, etc., which may negatively affect the resolution accuracy, thereby causing distortion of the final prediction result. In order to reduce the adverse effect of the above problems on the decomposition result, the EEMD adds white gaussian noise to the original signal and then decomposes the original signal, thereby obtaining good effect. However, due to the difference of artificially selected white noise, the eigenmode functions obtained by decomposition are also different, which makes the EEMD method unstable, and it is difficult for the EEMD method to completely eliminate the reconstruction error caused by the added gaussian white noise, which also leads to the increase of difficulty in establishing an accurate prediction model. On the basis, by adding white noise with the characteristics of self-adaptive characteristics and the like to the EMD, the scholars successively put forward complementary integrated empirical mode decomposition (CEEMDAN) and Improved CEEMDAN (ICEEMDAN), and the adverse effect of the modal aliasing problem on the precision of a decomposition result is gradually reduced. However, the improved algorithms such as EEMD still have the problem that it is unable to sufficiently decompose the close frequency components, so that the effect of decomposing high frequency data is limited.
VMD is a multi-resolution technology that originates from signal processing. Unlike EMD, VMD is a completely non-recursive algorithm that can decompose an original signal or sequence data into multiple components (subsequences) with specified bandwidths in the spectrum. Research shows that VMD is far superior to similar models in the aspects of noise suppression, decomposition precision improvement and the like. However, the number of modalities obtained by VMD decomposition needs to be preset, but there is no uniform guidance for presetting this value, so that the final decomposition and prediction result will be obviously affected. Too many decomposition modalities will cause the waste of computing resources, while too few decomposition modalities will cause more noise data in the modalities, resulting in a high difficulty of accurate prediction, and finally a low prediction accuracy.
Disclosure of Invention
The invention aims to provide a method for predicting the monthly stock keeping quantity of live pigs, which realizes accurate prediction of the monthly stock keeping quantity of the live pigs by introducing modal decomposition standard modified Variational Modal Decomposition (VMD) and combining the VMD with an Extreme Learning Machine (ELM) based on a decomposition-integration framework.
To solve the above technical problem, an embodiment of the present invention provides the following solutions:
a method for predicting monthly stock quantity of live pigs comprises the following steps:
decomposing original time sequence data x (t) of the live pig stock quantity into K modes by a variational mode decomposition algorithm;
for each mode obtained by the variational mode decomposition, normalization processing is carried out, and then the modes are divided into a training set and a test set according to a preset proportion;
training an extreme learning machine algorithm by using training set data, and determining optimal parameters of the algorithm;
selecting input set data of an extreme learning machine algorithm by using a sliding window with the step length of V;
for each mode, inputting the input set data of the test set into the trained extreme learning machine algorithm, outputting the predicted value of the next time point, and performing reverse normalization processing on the predicted value to obtain a predicted value sequence u k (t);
And adding and reconstructing the predicted values of all the modes to obtain a final predicted value result.
Preferably, K is according to the index r res Adaptively determined, the formula is as follows:
wherein x (t) represents the original time series, and N represents the total number of time points; when r is res When the number of modes is less than 0.01, the mode number K is determined.
Preferably, for each mode obtained by the variation mode decomposition, the formula for performing the normalization process is as follows:
preferably, the preset ratio is 8: 2.
preferably, the formula for the additive reconstruction of the predicted values of all modalities is as follows:
preferably, the process of the variation modal decomposition algorithm is as follows:
initializing the bandwidth u of the modal component k Center frequency ω k And a cycle number n;
when omega is greater than or equal to 0, circularly updating the bandwidth u of each modal component k And center frequencyω k Adaptively decomposed into a frequency at the center frequency ω k Modal components for central diffusion:
wherein u is k For the decomposed modal component, omega k And λ is a lagrange multiplier for the center frequency corresponding to the modal component, and the loop is terminated when the following condition is satisfied:
wherein epsilon is a constant;
when each component is calculated, distributing the signal into a variation model for decomposition, and realizing the decomposition of the complex signal by searching the optimal solution of a constraint variation model by using Hilbert transform and Gaussian smoothing; solving the hilbert transform of each mode, then shifting the spectrum of each mode to the baseband, then using H-gaussian smoothing of the demodulated signal, minimizing the sum of the mode bandwidths:
introducing a secondary penalty factor alpha and a Lagrange multiplier lambda, converting the constraint problem into a non-constraint problem, and solving by adopting an alternative multiplier direction method:
preferably, the extreme learning machine algorithm is implemented as follows:
given a sample (x) i ,y i ) Wherein x is i =[x i1 ,x 2 ,...,x iN ] T Representing an N-dimensional input set, y i =[y i1 ,y i2 ,...,y iM ] T Representing an M-dimensional output set, i ═ 1, 2., and N represents a sample label; the calculation process of the extreme learning machine algorithm is as follows:
wherein L represents the number of hidden layer nodes, beta l An output weight matrix representing the l-th node, F being an activation function for performing the operation, w l And b l Respectively representing an input weight vector and an offset vector of the ith node of the hidden layer; thus, the above operation may be expressed as H β ═ Y, where β ═ Y 1 ,β 2 ,...,β L ] T H denotes the hidden layer output weight matrix, expressed as follows:
thus, β is expressed as: beta is a T =H + T=H T (HH T ) -1 T, wherein H + =H T (HH T ) -1 A Moore-Penrose generalized inverse matrix representing the hidden layer output matrix H.
Preferably, the method further comprises:
and comparing the predicted value sequence with the actual value sequence to verify the prediction effect.
The technical scheme provided by the embodiment of the invention has the beneficial effects that at least:
according to the method, a modal decomposition standard modified Variational Modal Decomposition (VMD) algorithm is introduced, and the VMD algorithm is combined with an Extreme Learning Machine (ELM) algorithm on the basis of a decomposition-integration framework, so that the monthly stock quantity of the live pigs is accurately predicted.
Currently, in the variation modal decomposition, the modal number K needs to be preset, but the preset of the value lacks a uniform guidance, so that the final decomposition and prediction result are obviously influenced. Too many decomposition modalities will cause the waste of computing resources, while too few decomposition modalities will cause more noise data in the modalities, resulting in a high difficulty of accurate prediction, and finally a low prediction accuracy. The invention provides a decomposition standard suitable for complex time sequence data, so that the number K of the optimal decomposition modes can be determined adaptively by the variational mode decomposition under the standard, the calculation resources are saved, the obvious noise reduction of the original sequence is realized, and a foundation is laid for the prediction work of the later stage.
Considering the problems that the training speed is low, local optimization is easy to fall into, overfitting is easy to occur and the like when the traditional ANN adjusts the weight parameters by adopting a gradient descent iterative algorithm, and the SVR has high sensitivity on selection of a kernel function and parameters thereof in the aspect of generalization capability, the invention adopts the ELM with high operation speed and strong generalization capability as a main prediction model, thereby greatly improving the operation efficiency and the convergence speed and enabling the prediction result to be more robust.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a flowchart of a method for predicting monthly stock keeping amount of live pigs according to an embodiment of the present invention;
FIG. 2 is a schematic illustration of a single step prediction of future 1 order data;
FIG. 3 is a graphical representation of the comparison of predicted results between ELM, EMD-ELM, ICEEMDAN-ELM, and VMD-ELM.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
The embodiment of the invention provides a method for predicting monthly stock keeping amount of live pigs, which comprises the following steps as shown in figure 1:
the raw time series data x (t) of the live pig stock volume is decomposed into K modalities, also called eigenmode functions (IMF), by a Variational Modal Decomposition (VMD) algorithm.
Wherein K is according to the index r res Adaptively determined, the formula is as follows:
wherein x (t) represents the original time sequence, and N represents the total number of time points; when r is res The number of modes K can be determined when a level of less than 0.01 is reached and there is no significant downward trend.
For each mode obtained by Variable Mode Decomposition (VMD), normalization processing is carried out, and then the modes are divided into a training set and a test set according to a preset proportion.
The formula for normalization is as follows:
as a specific embodiment of the invention, after normalization processing, the training set and the test set are divided according to a ratio of 8: 2.
Training the extreme learning machine algorithm by using the training set data, determining the optimal parameters of the algorithm, and continuously and iteratively updating the parameters in the network in the training process.
The data of the first order in the future is predicted in a single step by adopting a sliding window with the step size V, namely, historical data of the V order is taken as input set data of an Extreme Learning Machine (ELM) algorithm, and the data of the first order in the future is predicted in a single step, as shown in FIG. 2.
For each mode, inputting the input set data of the test set into a trained Extreme Learning Machine (ELM) algorithm, outputting a predicted value of the next time point, and performing inverse normalization processing on the predicted value to obtain a predicted value sequence u k (t)。
Adding and reconstructing the predicted values of all the modes to obtain a final predicted value result, wherein the calculation formula is as follows:
further, the method further comprises:
and comparing the predicted value sequence with the actual value sequence to verify the prediction effect of the model.
Further, in an embodiment of the present invention, the process of the Variational Modal Decomposition (VMD) algorithm is as follows:
initializing the bandwidth u of the modal component k Center frequency ω k And a cycle number n;
when omega is greater than or equal to 0, circularly updating the bandwidth u of each modal component k And center frequency omega k Adaptively decomposed into a frequency at the center frequency ω k Modal components for central diffusion:
wherein u is k For the decomposed modal component, omega k Is the center frequency corresponding to the modal component, and λ is Lagrange multiplier, when the following conditions are satisfiedAnd, terminating the cycle:
wherein epsilon is a constant;
VMD is a novel multi-component signal decomposition algorithm, when calculating each component, the signal is distributed into a variation model for decomposition, Hilbert transform and Gaussian smoothing are used, and the decomposition of complex signals is realized by searching the optimal solution of a constraint variation model; solving the hilbert transform of each mode, then shifting the spectrum of each mode to the baseband, then minimizing the sum of the mode bandwidths using H-gaussian smoothing of the demodulated signal:
introducing a secondary penalty factor alpha and a Lagrange multiplier lambda, converting the constraint problem into a non-constraint problem, and solving by adopting an alternative multiplier direction method:
further, in the embodiment of the present invention, the implementation process of the Extreme Learning Machine (ELM) algorithm is as follows:
given a sample (x) i ,y i ) Wherein x is i =[x i1 ,x 2 ,...,x iN ] T Representing an N-dimensional input set, y i =[y i1 ,y i2 ,...,y iM ] T Representing an M-dimensional output set, i ═ 1, 2., and N represents a sample label; the calculation process of the extreme learning machine algorithm is as follows:
wherein L represents the number of hidden layer nodes, beta l An output weight matrix representing the l-th node, F being an activation function for performing the operation, w l And b l Respectively representing an input weight vector and an offset vector of the ith node of the hidden layer; thus, the above operation may be expressed as H β ═ Y, where β ═ Y 1 ,β 2 ,...,β L ] T H denotes the hidden layer output weight matrix, expressed as follows:
thus, β can be expressed as: beta is a T =H + T=H T (HH T ) -1 T, wherein H + =H T (HH T ) -1 A Moore-Penrose generalized inverse matrix representing the hidden layer output matrix H.
In order to verify the superiority of the mixed model VMD-ELM algorithm, the method is compared with other methods, including single prediction models SVR, BPNN and ELM, mixed prediction models EMD-SVR, EMD-BPNN and EMD-ELM, ICEEMDAN-SVR, ICEEMDAN-BPNN and ICEEMDAN-ELM, VMD-SVR and VMD-BPNN. All models adopt the same data set, the original live pig stock quantity data is directly used as an input data set of three single models, namely SVR, BPNN and ELM for model training, and the rest models are respectively trained by using data obtained by decomposition by EMD, ICEEMDAN and VMD methods. Table 1 shows the prediction error of the mixed model (VMD-ELM) and other models in monthly stock quantity data of the Chinese pigs.
TABLE 1 comparison of prediction errors for different prediction methods
As can be seen from Table 1, the four prediction error indexes MAE, RMSE, MAPE and TIC of the mixed model provided by the invention on the monthly stock quantity data of the live pigs are all significantly smaller than those of other models. The calculation formulas of the four evaluation criteria are as follows:
FIG. 3 further shows a comparison of predicted results between ELM, EMD-ELM, ICEEMDAN-ELM and VMD-ELM. The result shows that the prediction result of the single prediction model ELM has a remarkable 'lag prediction' effect, namely the predicted value of the future stage is easily influenced by the last stage observation value, so that the fact that the predicted value sequence curve is similar to the current observation value curve obtained by shifting the current observation value curve backwards for the first stage is shown, and finally the result of relatively large error level of the single prediction model is caused. The reason for this phenomenon is that the monthly stock data of the live pigs have typical nonlinear and non-stationary characteristics, and a large amount of noise exists in the sequence, so that the model training is difficult and the generalization capability is relatively poor. A test set sample of this data was data from 7 months in 2019 to 11 months in 2021, including monthly data for 29 months, which originated from the chinese government network (http:// www.gov.cn) and the look-ahead database (d.qaanzhan. Wherein the data of 2021 year is official forecast data. Experimental results show that the VMD-ELM mixed model prediction method provided by the invention can obtain higher prediction precision in the prediction of monthly stock quantity sequence data of Chinese pigs.
In conclusion, the invention provides an adaptive prediction method based on Variational Modal Decomposition (VMD) and Extreme Learning Machine (ELM) for accurately predicting the monthly stock quantity of live pigs.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (7)
1. A method for predicting monthly stock quantity of live pigs is characterized by comprising the following steps:
decomposing original time sequence data x (t) of the live pig stock quantity into K modes by a variational mode decomposition algorithm;
for each mode obtained by the variational mode decomposition, normalization processing is carried out, and then the modes are divided into a training set and a test set according to a preset proportion;
training an extreme learning machine algorithm by using training set data, and determining optimal parameters of the algorithm;
selecting input set data of an extreme learning machine algorithm by using a sliding window with the step length of V;
for each mode, inputting the input set data of the test set into the trained extreme learning machine algorithm, outputting the predicted value of the next time point, and performing reverse normalization processing on the predicted value to obtain a predicted value sequence u k (t);
Adding and reconstructing the predicted values of all the modes to obtain a final predicted value result;
the implementation process of the extreme learning machine algorithm is as follows:
given a sample (x) i ,y i ) Wherein x is i =[x i1 ,x 2 ,...,x iN ] T Representing an N-dimensional input set, y i =[y i1 ,y i2 ,...,y iM ] T Representing an M-dimensional output set, i ═ 1, 2., and N represents a sample label; the calculation process of the extreme learning machine algorithm is as follows:
wherein L represents the number of hidden layer nodes, beta l An output weight matrix representing the l-th node, F is an activation function for performing the operation, w l And b l Respectively representing an input weight vector and an offset vector of the ith node of the hidden layer; the above calculation process can therefore be described again as H β ═ Y, where β ═ Y 1 ,β 2 ,...,β L ] T H denotes the hidden layer output weight matrix, expressed as follows:
thus, β is expressed as: beta is a T =H + T=H T (HH T ) -1 T, wherein H + =H T (HH T ) -1 A Moore-Penrose generalized inverse matrix representing the hidden layer output matrix H.
2. The method for predicting monthly stock keeping quantity of live pigs according to claim 1, wherein K is determined according to an index r res Adaptively determined, the formula is as follows:
wherein x (t) represents the original time series, N represents the totalCounting the number of the time points; when r is res When the number of modes is less than 0.01, the mode number K is determined.
4. the method for predicting monthly stock keeping quantity of live pigs according to claim 1, wherein the preset ratio is 8: 2.
6. the method for predicting monthly stock inventory of live pigs according to claim 1, wherein the process of the variational modal decomposition algorithm is as follows:
initializing the bandwidth u of the modal component k Center frequency ω k And a cycle number n;
when omega is greater than or equal to 0, circularly updating the bandwidth u of each modal component k And center frequency omega k Adaptively decomposed into a frequency at the center frequency ω k Modal components for central diffusion:
wherein u is k For the decomposed modal component, omega k And λ is a lagrange multiplier for the center frequency corresponding to the modal component, and the loop is terminated when the following condition is satisfied:
wherein epsilon is a constant;
when each component is calculated, distributing the signal into a variation model for decomposition, and realizing the decomposition of the complex signal by searching the optimal solution of a constraint variation model by using Hilbert transform and Gaussian smoothing; solving the hilbert transform of each mode, then shifting the spectrum of each mode to the baseband, then minimizing the sum of the mode bandwidths using H-gaussian smoothing of the demodulated signal:
introducing a secondary penalty factor alpha and a Lagrange multiplier lambda, converting the constraint problem into a non-constraint problem, and solving by adopting an alternative multiplier direction method:
7. the method of predicting monthly stock inventory of live pigs according to claim 1, further comprising:
and comparing the predicted value sequence with the actual value sequence to verify the prediction effect.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110339114.9A CN113011669B (en) | 2021-03-30 | 2021-03-30 | Method for predicting monthly stock quantity of live pigs |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110339114.9A CN113011669B (en) | 2021-03-30 | 2021-03-30 | Method for predicting monthly stock quantity of live pigs |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113011669A CN113011669A (en) | 2021-06-22 |
CN113011669B true CN113011669B (en) | 2022-08-02 |
Family
ID=76409205
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110339114.9A Expired - Fee Related CN113011669B (en) | 2021-03-30 | 2021-03-30 | Method for predicting monthly stock quantity of live pigs |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113011669B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114037130A (en) * | 2021-10-28 | 2022-02-11 | 重庆川南环保科技有限公司 | Household appliance maintenance order prediction method and device and terminal equipment |
CN115438301B (en) * | 2022-10-26 | 2023-06-20 | 北京建筑大学 | Equipment operation trend prediction method based on ICEEMDAN (information and communication technology) secondary decoupling index model |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106897794A (en) * | 2017-01-12 | 2017-06-27 | 长沙理工大学 | A kind of wind speed forecasting method based on complete overall experience mode decomposition and extreme learning machine |
CN109299430A (en) * | 2018-09-30 | 2019-02-01 | 淮阴工学院 | The short-term wind speed forecasting method with extreme learning machine is decomposed based on two stages |
CN109816236A (en) * | 2019-01-18 | 2019-05-28 | 中国农业科学院农业信息研究所 | A kind of planing method and planning system of pig-breeding |
CN111193256A (en) * | 2020-01-07 | 2020-05-22 | 上海电力大学 | Power load prediction method based on variational modal decomposition and gated cyclic unit |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190188611A1 (en) * | 2017-12-14 | 2019-06-20 | Business Objects Software Limited | Multi-step time series forecasting with residual learning |
-
2021
- 2021-03-30 CN CN202110339114.9A patent/CN113011669B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106897794A (en) * | 2017-01-12 | 2017-06-27 | 长沙理工大学 | A kind of wind speed forecasting method based on complete overall experience mode decomposition and extreme learning machine |
CN109299430A (en) * | 2018-09-30 | 2019-02-01 | 淮阴工学院 | The short-term wind speed forecasting method with extreme learning machine is decomposed based on two stages |
CN109816236A (en) * | 2019-01-18 | 2019-05-28 | 中国农业科学院农业信息研究所 | A kind of planing method and planning system of pig-breeding |
CN111193256A (en) * | 2020-01-07 | 2020-05-22 | 上海电力大学 | Power load prediction method based on variational modal decomposition and gated cyclic unit |
Non-Patent Citations (1)
Title |
---|
基于HP分离ARIMA—M...v模型的我国生猪存栏量预测;吴雪;《统计与决策》;20130630(第6期);第102-104页 * |
Also Published As
Publication number | Publication date |
---|---|
CN113011669A (en) | 2021-06-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | Smart wind speed forecasting approach using various boosting algorithms, big multi-step forecasting strategy | |
CN113011669B (en) | Method for predicting monthly stock quantity of live pigs | |
Huang et al. | A hybrid model for short-term dissolved oxygen content prediction | |
CN112884236B (en) | Short-term load prediction method and system based on VDM decomposition and LSTM improvement | |
CN112668611A (en) | Short-term photovoltaic power generation power prediction method based on Kmeans and CEEMD-PE-LSTM | |
CN111353534B (en) | Graph data category prediction method based on adaptive fractional order gradient | |
CN113836823A (en) | Load combination prediction method based on load decomposition and optimized bidirectional long-short term memory network | |
CN114970345A (en) | Short-term load prediction model construction method, device, equipment and readable storage medium | |
CN115344927A (en) | High arch dam construction simulation parameter dynamic updating method based on engineering historical data | |
Sun et al. | Spatiotemporal wind power forecasting approach based on multi-factor extraction method and an indirect strategy | |
CN114676645A (en) | Non-stationary time sequence prediction method and system | |
CN113222234B (en) | Gas demand prediction method and system based on integrated modal decomposition | |
Sun et al. | Multi-step wind speed forecasting based on a hybrid decomposition technique and an improved back-propagation neural network | |
Xinxin et al. | Short-term wind speed forecasting based on a hybrid model of ICEEMDAN, MFE, LSTM and informer | |
Li et al. | Short-term wind power prediction based on modal reconstruction and CNN-BiLSTM | |
Liu et al. | A wind speed forcasting model based on rime optimization based VMD and multi-headed self-attention-LSTM | |
Ao et al. | Entropy estimation via normalizing flow | |
CN113128666A (en) | Mo-S-LSTMs model-based time series multi-step prediction method | |
CN112232565A (en) | Two-stage time sequence prediction method, prediction system, terminal and medium | |
CN115619563A (en) | Stock price analysis method based on neural network | |
CN115081323A (en) | Method for solving multi-objective constrained optimization problem and storage medium thereof | |
Zhang et al. | A novel hybrid forecasting system based on data augmentation and deep learning neural network for short-term wind speed forecasting | |
CN114722490A (en) | Agent model global optimization method based on mixed increase and interval reduction | |
Zhang et al. | A Novel Combined Model Based on Hybrid Data Decomposition, MSWOA and ENN for Short-Term Wind Speed Forecasting | |
Liu et al. | Predicting sunspot numbers based on inverse number and intelligent fixed point |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220802 |